Public Service Announcement from RH on fragile systems

I recently read Taleb's Antifragile.  Let's ignore the shrill insistence on the idea that he doesn't need an editor. This is a claim that sits on a spectrum somewhere between self-deluding and narcissistically demented because he really does need one.  The message underneath what needs editing, however, is solid and has vast applicability to the retirement project.  I was going to write a post on this once but FarnamStreet did one better and sooner.  But that is not my point here.  My point here is that I just went through a knock-down hard-drive crash that, while I have sorta-recent backups, is more or less going to force me to walk away from a track record of 8-9 years of my systematic rules-based investing that shows that I am a genius (a genius like T doesn't need an editor. Follow me? for the dense: I am not a genius...though the track record was, in fact, world class imho).  In a parallel universe, this release from the bonds of continuing to do what I was doing is probably a big fat relief.  In this universe it is a pain especially since I had just read Antifragile where one of the main points of the book (among many) is to brook no risk from single points of failure.  I had a big SPOF and I even knew it before it bit me. That thought bites me.  I now have a disc going to Kentucky for a level 2 data extraction.  Probably at a CIA black ops site. 4-6 weeks.  $400. Nothing guaranteed.

So here is my RH PSA:

Don't allow single points of failure

Back-up more frequently if not instantaneously

Don't build highly custom software that depends on a particular platform and a set of machines

Don't assume machines live forever

Pay attention to your power supply and its continuity

Don't get attached to really old commercial software that is so out of date that it probably pre-dates disco.

If you instantiate 3 years of "apex" knowledge capital into software just before a procrastinated backup cycle then :'-(

Custom-coded, mirrored, amateur RAID arrays are really cute only if the process backs up very often

Don't allow single points of failure